[MIII2MII3]n+ trigonal bipyramidal cages based on diamagnetic and paramagnetic metalloligands

A family of [MIII2MII3]n+ trigonal bipyramidal cages are characterised in the solution and solid state.


Introduction
Molecular magnetism relies on the ability of the synthetic chemist to make an enormous breadth of structurally diverse polymetallic cages spanning the d-and f-block of the periodic table. 1-10 The structural and magnetic characterisation of such species details the magneto-structural relationship and oen uncovers fascinating magnetic phenomena which, in turn, feedback into the synthesis of new complexes designed to enhance and improve properties toward application. [11][12][13][14][15][16][17][18] Synthetic strategies for the design of polymetallic clusters containing multiple paramagnetic metal ions span the range from serendipitous self-assembly in which coordinatively exible metal ions, that can oen exist in multiple oxidation states, are combined with organic ligands capable of bridging in numerous ways to form complexes whose absolute structures are difficult to predict, through to a more 'supramolecular' approach whereby metal ions with dened coordination geometries are paired with rigid ligands containing donor atoms with a single, predesigned orientation preference that afford, in most cases, a predicted structure. In the eld of molecular magnetism, the latter is perhaps best exemplied by cyanometalate chemistry. [19][20][21][22][23] A similar synthetic approach is followed in the metallosupramolecular chemistry of diamagnetic cages and capsules where the combination of directional metal-ligand bonding and rigorously rigid ligands creates cages with permanent internal cavities capable of hosting guest molecules, constructed primarily for potential application in, for example, catalysis, 24 the stabilisation of reactive molecules 25 and photochemistry. 26 Due to the difficulties associated with performing solution-based one-and two-dimensional NMR spectroscopy on paramagnetic species, where broad signals and a wide chemical shi range are commonplace, 27 it is perhaps not surprising that the majority of metallosupramolecular chemistry has focused on the use of diamagnetic metal centres, albeit with some notable exceptions. 28 We recently initiated a project that would enable heterometallic, paramagnetic coordination cages to be accessed in a modular and predictable fashion, 29 an approach centred around the tritopic metalloligand [M III L 3 ] (where HL ¼ 1-(4pyridyl)butane-1,3-dione), which features a tris(acac) octahedral transition metal core functionalised with three p-pyridyl donor groups (Fig. 1). 30

Experimental section
Syntheses 1-(4-Pyridyl)butane-1,3-dione (HL) and the metalloligand [Cr III L 3 ] were prepared according to previously published procedures. 29, 35 All reactions were performed under aerobic conditions. Solvents and reagents were used as received from commercial suppliers. Elemental analyses were carried by Medac Ltd.

Crystal structure information
For compounds 1, 2 and 3 single-crystal X-ray diffraction data were collected at T ¼ 100 K on a Rigaku AFC12 goniometer equipped with an enhanced sensitivity (HG) Saturn 724+ detector mounted at the window of an FR-E+ Superbright MoKa rotating anode generator with HF Varimax optics (70 mm focus) 36 using Rigaku Crystal Clear and CrysalisPro soware 37,38 for data collection and reduction. The crystals were sensitive to solvent loss and were therefore 'cold-mounted' using X-Temp 2 System apparatus at T ¼ 70 C and then quickly transferred to diffractometer.
For compounds 4 and 5 single crystal X-ray diffraction data were measured on a Rigaku Oxford Diffraction SuperNova diffractometer using Cu radiation at T ¼ 120 K. The CrysalisPro soware package was used for instrument control, unit cell determination and data reduction. 39 Unit cell parameters in all cases were rened against all data. Crystal structures were solved using the charge ipping method implemented in SUPERFLIP 40 (1, 2, and 3), or by direct methods with ShelXS (4 and 5). All structures were rened on F o 2 by full-matrix leastsquares renements using ShelXL 41 within the OLEX2 suite. 42 All non-hydrogen atoms were rened with anisotropic displacement parameters, and all hydrogen atoms were added at calculated positions and rened using a riding model with isotropic displacement parameters based on the equivalent isotropic displacement parameter (U eq. ) of the parent atom. All ve structures contain accessible voids and channels that are lled with diffuse electron density belonging to uncoordinated solvent, and CF 3 SO 3 À anions in the case of compounds 4-5. The SQUEEZE routine of PLATON 43 was used to remove remaining electron density corresponding to solvent and anions not reported in the calculated formula. Crystallographic summary and structure renement details are presented in Table 1. CCDC: 1520425-1520429. †

Physical measurements
Magnetisation measurements were carried out on a Quantum Design SQUID MPMS-XL magnetometer, operating between 1.8 and 300 K for DC applied magnetic elds ranging from 0 to 5 T. Microcrystalline samples were dispersed in eicosane in order to avoid torquing of the crystallites. Heat capacity measurements were carried out for temperatures down to ca. 0.3 K by using a Quantum Design 9T-PPMS, equipped with a 3 He cryostat. The experiments were performed on thin pressed pellets (ca. 1 mg) of a polycrystalline sample, thermalised by ca. 0.2 mg of Apiezon N grease, whose contribution was subtracted by using a phenomenological expression. X-and Q-band EPR spectra were collected on powdered microcrystalline samples of [FeL 3 ] and compounds 1-4 at the UK National EPR Facility in Manchester.

Results and discussion
Solution self-assembly and structure We were thus pleased to nd that when we switched to the more strongly coordinating bis(diphenylphosphino)propane (dppp), we were able to isolate the [Al 2 Pd 3 ] 6+ trigonal bipyramidal complex, 5, in 83% yield following reaction overnight at 50 C between [Al III L 3 ] and [Pd(dppp)(OTf) 2 ] in acetonitrile. All the spectroscopic data indicate that the structure of 5, conrmed by X-ray crystallography (see below), is preserved in solution. As well as ESI-MS, which reveals the 3+ charge state corresponding to [5 À 3OTf] 3+ matching the expected isotopic distribution (see ESI †), the 1 H NMR spectrum of the product (Fig. 2b) shows just a single set of signals. The 1 H DOSY spectrum also indicates that all the resonances possess the same diffusion coefficient, which corresponds to a hydrodynamic radius of 9.9Å, closely matching the data obtained by XRD. It is also interesting to note that the starting metalloligand [Al III L 3 ] exists as a mixture of the mer and fac congurations, clearly evidenced by the multiplet for the acac CH and CH 3 signals in the 1 H NMR spectrum (Fig. 2c, resonances shown in blue and magenta), which is replaced by a singlet in the crude reaction mixture (Fig. 2d). This indicates that under the conditions of the reaction, [Al III L 3 ] is congurationally dynamic, and that the self-assembly process amplies the proportion of the fac conguration through the formation of 5. While mer tris(bidentate) octahedral complexes are also known to generate discrete metallosupramolecular cages, 45 the divergent disposition of the pendant donor groups create larger closed systems, which with a dynamic system such as this will rapidly rearrange to give the entropically more favourable trigonal bipyramid. A comparison of the 1 H NMR spectra of the re-dissolved crystalline sample of 5 (Fig. 2b) and the crude reaction solution, obtained by treating a slight excess of [Pd(dppp)(OTf) 2 ] with [Al III L 3 ] in CD 3 CN (Fig. 2d), shows that this amplication is not a solid-state packing effect, rather a solution-based effect. The single set of signals in the 1 H NMR spectrum of the product (Fig. 2b/d) also indicates that 5 is formed with complete diastereoselectivity. 46 This represents a second tier of self-sorting, which, unusually, involves Pd-mediated heterochiral recognition of D and L-[Al III L 3 ] enantiomers (see below).

Solid-state structure descriptions
The heterometallic trigonal bipyramid cages [Fe 2 Co 3 L 6 Cl 6 ] (1), [Fe 2 Zn 3 L 6 Br 6 ] (2), [Cr 2 Zn 3 L 6 Br 6 ] (3), [Cr 2 Pd 3 L 6 (dppp) 3 ](OTf) 6 (4) and [Al 2 Pd 3 L 6 (dppp) 3 ](OTf) 6 (5) (Fig. 3 and 4) were all synthesised in a similar manner, by addition of either tetrahedral or cis-protected square planar M II compounds to the metal-   While the intrametallic distances of the ve trigonal bipyramids are similar, there is nonetheless a distinct diastereomeric difference between structures 1-3 and 4-5. Whereas 1-3 are all homochiral racemates in which each intact capsule features two [M III L 3 ] units that possess the same L or D chirality, in contrast structures 4 and 5 are both the achiral heterodiastereomer. While sorting of chiral octahedral metal motifs has been frequently observed in metallosupramolecular assembly reactions, for the vast majority homochiral assemblies are energetically preferred. 47 The commonality of the [Pd(dppp)] unit in both 4 and 5 that feature different [M III L 3 ] metalloligands would suggest that either the small change in angle between pyridine donors at each M II connector and/or the interactions of the dppp protecting ligand with these donors cause the change in diastereomeric preference. Solution studies with 5 would also indicate this is not simply due to selective crystallization from a complex mixture (see above). Outwith cyanometalate

SQUID magnetometry
The dc (direct current) molar magnetic susceptibility, c, of a polycrystalline sample of 1 was measured in an applied magnetic eld, B, of 0.1 T, over the 2-300 K temperature, T, range. The experimental results are shown in Fig. 5 in the form of the cT product, where c ¼ M/B, and M is the magnetisation of the sample. At room temperature, the cT product of 1 has a value of 14.4 cm 3 K mol À1 , in good agreement with the sum of Curie constants for a [Fe III 2 Co II 3 ] unit (14.375 cm 3 K mol À1 , g Fe ¼ g Co ¼ 2.0). Note that the estimation of the g-value of the Co II ions here is an approximation and subject to error (e.g. lattice solvent lost upon sample drying will result in a variation of the samples diamagnetism), and a better measure comes from the EPR spectroscopy, which is consistent with g Co ¼ 2.3 (vide infra). Upon cooling, the cT product of 1 remains essentially constant down to approximately 100 K, wherefrom it decreases upon further cooling to 9.5 cm 3 K mol À1 at 2 K. Given that the anisotropy of Fe III is negligible, this behaviour is consistent with a relatively large single-ion magnetic anisotropy for the Co II centres and/or an antiferromagnetic exchange interaction between the Fe III and Co II centres. To better dene the lowtemperature magnetic properties of 1, low temperature variable-temperature-and-variable-eld (VTVB) magnetisation data were measured in the temperature and magnetic eld ranges T ¼ 2-12 K and B ¼ 0-5 T (Fig. 5). At the highest investigated eld (5 T) and the lowest investigated temperature (2 K), the magnetisation of 1 is of 13.7 m B (m B is the Bohr magneton). Furthermore, when the VTVB data of 1 are plotted against the reduced quantity m B B/kT, little nesting of the VTVB data is observed. This observation indicates that the part of the energy spectrum of 1 probed under these experimental conditions does not present signicant anisotropy splitting with respect to the temperature of measurement at zero magnetic eld.
For the quantitative interpretation of the magnetisation data, we used spin-Hamiltonian (1) (1) where the summation indexes i, j run through the constitutive metal centres, g i is the g-factor of the i th centre,Ŝ is a spin operator, J is the isotropic exchange interaction parameter, D is the uniaxial anisotropy parameter and S is the total spin. In our spin-Hamiltonian model, we assume for simplicity that all g-factors are equal to 2, S Fe III ¼ 5/2, S Co II ¼ 3/2, we only consider exchange interactions between Co II and Fe III centres, and neglect the single-ion anisotropy of Fe III . Furthermore, we x the uniaxial anisotropy of Co II to D Co ¼ À14 cm À1 , as extracted from the modelling of the EPR data and theoretical calculations, which are discussed further in the following sections. Thus, at this point our model contains only one free parameter, namely, the isotropic exchange between Fe III and Co II , J Fe-Co . The cT product of 1 was tted to spin-Hamiltonian (1) by full matrix numerical diagonalisation of the spin-Hamiltonian of the full system of dimension 2304 by 2304, through use of the Levenberg-Marquardt algorithm. 53 This resulted in the best-t parameter J Fe-Co ¼ À0.04 cm À1 . In order to verify the validity of our model, J Fe-Co was xed to the determined best-t value, J Fe-Co ¼ À0.04 cm À1 , and D Co was maintained xed at À14 cm À1 . At this point our model contains no free parameters. Thereaer, the VTVB data of 1 were simulated by use of spin-Hamiltonian (1). The simulated curves are shown as solid red lines in Fig. 5. With these parameters, the energy spectrum of 1 consists of four groups of densely packed states, each separated by approximately 2D Co (Fig. 6). It is interesting to note that multiple ground level crossings simultaneously occur at approximately 0.47 T when the magnetic eld is applied parallel to the quantisation axis.
Heat capacity Fig. 7 shows the collected heat capacity data, normalised to the gas constant, c p /R of 1 as a function of temperature (between ca. 0.3 K and 30 K) for zero-applied magnetic eld. As is typical for molecular magnetic materials, 54 lattice vibrations contribute predominantly to c p as a rapid increase above liquid-helium temperature. The lattice contribution can be described by the Debye model (dotted line in Fig. 7), which simplies to a c p /R ¼ aT 3 dependence at the lowest temperatures, where a ¼ 7.6 Â 10 À3 K À3 for 1.
For T < ca. 3 K, the zero-eld c p shows a wide bump-like anomaly, which we attribute to the splitting of the spin levels by zero-eld splitting and magnetic interactions. At such low temperatures, the magnetic measurements are very sensitive to the applied magnetic eld, as seen in the experimental behaviour for elds of 3 T and higher (inset of Fig. 7). Such large intensities of the applied magnetic eld are sufficient for promoting full decoupling between the individual spin centres (we recall that the exchange interaction is as small as J Fe-Co ¼ À0.04 cm À1 on the basis of the t of the magnetometry data). Therefore, the temperature and eld dependence of the c p data in Fig. 7 (inset), collected for B $ 3 T, are particularly suitable for probing the inuence of crystal elds on 1, down to temperatures signicantly lower than the ones obtained in the magnetisation measurements.
The solid lines in Fig. 7 are the curves calculated for Hamiltonian (1), using the best-t parameters from the magnetothermal and spectroscopic data and theoretical calculations, i.e., D Co ¼ À14 cm À1 and the here-negligible J Fe-Co ¼ À0.04 cm À1 . The agreement with the experimental data is good, though not outstanding. Anticipating the discussion on the EPR spectra (vide infra), we have checked that adding a zero-eld splitting (ZFS) of D Fe ¼ À0.2 cm À1 at the Fe III sites does not improve the t. The discrepancy is most evident below ca. 1 K, where the experimental data have lower values than the calculated ones. This behaviour can be explained by a wider broadening of the low-lying energy spectrum, likely induced by higher-order anisotropy terms, which are not taken into account in Hamiltonian (1).

EPR spectroscopy
We previously reported EPR spectra of [CrL 3 ], giving the ZFS of the Cr III , s ¼ 3/2 ion as D ¼ À0.55 cm À1 with a small rhombicity of |E/D| ¼ 0.045. 29 Q-Band spectra of 3 and 4 are similar to that of [Cr III L 3 ], and give D ¼ À0.64 and À0.61 cm À1 , respectively (Fig. S11; † |E/D| ¼ 0.03-0.04). 55 Hence, the distortion imposed on the {CrO 6 } coordination sphere of [Cr III L 3 ] by complexation in the {Cr III 2 M II 3 } supramolecules results in a small, but measurable, increase of the ZFS at Cr III . The {CrO 6 } metric parameters do not appear to be very different.
Such an increase in D is also found for the Fe III (s ¼ 5/2) systems. X-and Q-band EPR spectra of [Fe III L 3 ] reveal a rather small ZFS of D ¼ 0.08 cm À1 with |E/D| ¼ 1/3 ( Fig. 8 and S12; † note the sign of D has no signicance with a fully rhombic Dtensor). These values are similar to those reported for [Fe(acac) 3 ] 57 On incorporation into the {Fe III 2 Zn II 3 } complex 2, a much richer spectrum is observed ( Fig. 8 and S12 7 Temperature dependence of the zero-field heat capacity c p , normalised to the gas constant R, for a polycrystalline sample of 1. The dotted line is the lattice contribution. Inset: temperature dependence of c p /R of 1 for T < 2 K and B $ 3 T. Solid lines are the best-fit curves, see text for details. fall within the observed features. The spectra also show that the J FeCo exchange interaction must be very weak, resulting only in severe broadening of the peaks. Test calculations on a simple {Fe III Co II } model, with xed ZFS at the s ¼ 5/2 and 3/2 spins (the latter taking D ¼ À14 cm À1 with E/D ¼ 0.1; averaging the results of CASSCF calculationssee below) suggest that if |J FeCo | > ca. 0.02 cm À1 then additional features would be observed in the Qband EPR spectrum. Note that the limit for the full, ve-spin system would be different.
The D Fe values obtained from EPR would have a negligible effect on the calculated cT(T) and c p (T,B) curves for 1, and a negligible effect on the global level structure in Fig. 6a, because both |D Fe | and |J FeCo | are (|D Co |. However, it would affect the detail of the states within each of the densely packed multiplets of Fig. 6a, because |D Fe | and |J FeCo | are of similar magnitude.

Theoretical studies
In order to independently verify the large ZFS of Co II we have performed complete active space self-consistent eld (CASSCF) calculations on the three unique Co II sites of 1, see the SI for details. The results suggest D Co ¼ À14 cm À1 , E/D ¼ 0.1 (Table S1 †) which is entirely consistent with the magnetometry and heat capacity data. The calculations also suggest that the principal axes of the local ZFS tensors are oriented roughly perpendicular to the Fe III -Fe III axis and canted approximately 120 with respect to one another in the plane (Fig. 9). Accounting for the non-collinearity in spin-Hamiltonian (1) did not improve the quality of the ts to the magnetometry or heat capacity data.

Conclusions
Complexes 1-5 represent a novel, and unusual family of trigonal bipyramidal cage complexes, built with the tritopic [ML 3 ] metalloligand, featuring a tris(acac) octahedral transition metal core functionalised with three p-pyridyl donor groups, and a series of transition metal salts. Outwith cyanometalate chemistry, compound 1 represents the rst example of such a cage containing paramagnetic metal ions. Complementary studies investigating the diamagnetic variants using 1 H NMR spectroscopy reveal some interesting features about the solution selfassembly process. Firstly, the [M III L 3 ] metalloligand is a highly dynamic tritopic building block as evidenced by fac congurational isomer being amplied at the expense of the mer during the course of cage formation. The self-assembly process also occurs with high and unusual stereoselectivity wherein the trigonal bipyramids are formed exclusively from twisted pyramidal components of opposite D/L-handedness. Solution stability of the cage is also conrmed via mass spectrometry. SQUID magnetometry and heat capacity measurements on 1 reveal weak antiferromagnetic exchange between the Fe III and Co II ions, with |D Co | ¼ 14 cm À1 . EPR spectroscopy reveals that the distortion imposed on the {MO 6 } coordination sphere of [M III L 3 ] by complexation in the {M III 2 M II 3 } supramolecules results in a small, but measurable, increase of the zero eld splitting at M III . CASSCF calculations on the three unique Co II sites of 1 suggest that the principal axes of the local ZFS tensors are oriented perpendicular to the Fe III -Fe III axis, but canted $120 with respect to each other.  Orientation of the principal anisotropy axis for the Co II sites in 1 (yellow rods); orange ¼ Fe, pink ¼ Co, green ¼ Cl, red ¼ O, blue ¼ N, beige ¼ C, white ¼ H.